Audit evidence preparation
Retrieve the fixed change-ticket sample, extract evidence for the same three test criteria, prepare source-linked workpapers, independently verify evidence and reach the auditor's conclusion. Same sample, audit coverage and evidence standard at both endpoints.
Evidence: Direct SOX extraction/annotation case supplies annual hours without a denominator; recent human-reviewed audit-document case reports fifty percent less processing time. Exact ticket-workflow endpoints remain conditional analyst assumptions.
Reference task model
The reference workload starts with 100 manual staff-hours. Organization settings adjust the hours below at the same work volume. Task shares and case budgets are assumptions. Checking, corrections and manual fallback are included.
| Operating work | Before staff-hours | After staff-hours fast–constrained | Saved staff-hours constrained–fast |
|---|---|---|---|
| Locate tickets and gather supporting evidence | 20 | 10–15 | 5–10 |
| Extract approval, testing and date attributes; populate rule checks | 30 | 15 | 15 |
| Draft and edit source-linked workpapers and annotations | 20 | 10–12 | 8–10 |
Independently verify original evidence, resolve exceptions and reach the conclusionAssumption and referencesAll thirty baseline auditor-review hours remain at the upper endpoint. Lower adds three hours to resolve omitted/conflicting attributes and annotations. Prepopulated pass/fail fields do not determine the final audit conclusion. Sources: 3 | 30 | 30–33 | -3–0 |
| Reference workload total | 100 | 65–75 | 25–35 |
Hours refer to this reference workload. Saved = before − after; negative saved hours add work. The fast case has fewer remaining hours. 25–35% saved before outward rounding to 25–35%. Displayed hours are rounded to 0.1. These are conditional scenarios, not confidence limits.
Reference scenario conditions
Constrained case · 25% saved · 75 hours remain
Partly indexed evidence and substantial editing leave 15 retrieval + 15 extraction + 12 drafting + 33 independent-review/correction hours: 75 remain, 25 saved.
Fast case · 35% saved · 65 hours remain
Indexed evidence, consistent formats and reusable workpapers leave 10 retrieval + 15 extraction + 10 drafting + 30 independent-review hours: 65 remain, 35 saved.
Other task mixes and faster examples
These additional assumptions show how work mix and coverage change the result. They are separate scenarios, not measured bounds or additions to the main saving.
Near-automatic evidence preparation
From 100 baseline human-hours, 35 remain and 65 are saved.
Complete indexed evidence and reliable template population reduce preparation from 70 to 5 hours. All 30 independent-review hours remain. This is an unvalidated automation scenario, beyond the quantified case reference.
Citations
Microsoft Frontier Finance: SOX Change Management Testing Agent
First-party internal deployment case · File modified 2026-05-07; public by 2026-05-11 according to completed source review
850 annual hours saved. Two agents extract ticket attributes, prepopulate rule checks and mark the matching source evidence.
Evidence period: Deployment/measurement dates and model version absent; listed technology includes Bing Chat. A 2026 file date does not establish recent deployment technology.
- No baseline hours, ticket volume, correction rate or timed sample. Cannot infer a percentage or establish current-model speed. Auditor verification remains in the scenario.
Provectus: Johnson Lambert reinvents insurance auditing with generative AI
Vendor-authored named customer deployment · 2026-04-20, first-party metadata verified in completed source review
50% lower document-processing time; auditors inspect extracted tables against originals and correct values.
Evidence period: Exact measurement period absent; listed Claude 3.7 Sonnet and Gemini 2.5 Pro establish post-2024 technology.
- PDF financial-report extraction differs from ticket testing. No timed sample or task breakdown. Other broader audit headlines conflict; do not use them or convert an efficiency increase into an equal time reduction. Its 50% document-processing result is neither a universal cap nor a guaranteed stage effect.
Netherlands AFM: 12 Building blocks for controlled use of auditing tools
Regulator guidance · 2025-12
No productivity result. Requires reliable input, verifiable outputs and continuing auditor responsibility.
Evidence period: 2025 supervisory guidance; older firm-selection data are unused.
- No numerical effect; none used.
How to interpret this range
Conditional planning scenarios, not measured population bounds or confidence intervals.
The same work is completed manually using existing systems and approved inputs. Sources and calculation models are usable. These are workflow reference cases, not company types.
- The workflow is in steady-state use and passes checks for equivalent output quality.
- Volume, coverage and output remain unchanged. Saved capacity is not assumed to reduce staffing.
- Human review, corrections, exceptions and existing approvals are included.
- Baseline task shares and transferred task-saving rates are analyst assumptions, not measured industry allocations.
Savings include conventional extraction, matching, retrieval and integration delivered with each scoped project. The separate marginal contribution of GenAI is not established.
Separate costs: Implementation and training, Separate project/platform upkeep, Upstream remediation outside the project, Additional output and newly worthwhile work.
Already-automated work, low adoption, difficult exceptions or unreliable output can produce smaller savings, zero or additional work. The positive lower endpoint is not a guaranteed minimum.
- Baseline task hours, eligible coverage and remaining human-work budgets are scenario assumptions.
- Published task results inform the scenario; no source establishes the complete workflow range.
- Same work volume and required output; manual fallback and corrections are included. Separate upkeep is excluded.
Convert to monthly hours: Monthly operating hours saved = baseline monthly workflow hours × saving percentage. Net capacity released = operating hours saved − annual project upkeep / 12 − allocated shared upkeep / 12. Shared upkeep is allocated once across the portfolio.
What would improve the estimate
- Direct SOX extraction/annotation case supplies annual hours without a denominator; recent human-reviewed audit-document case reports fifty percent less processing time. Exact ticket-workflow endpoints remain conditional analyst assumptions.
- Time matched manual and assisted preparation on the same fifty historical cases, with held-out adjudication
- Record attribute omissions, wrong evidence links, false passes and missing/conflicting approvals separately
- Measure preparer and independent auditor labor including correction and source re-performance; retain the same evidential standard
- Test whether preparation fits the reference 35–42 hour budget per 100 baseline hours, while preserving the full auditor-review and correction effort.
Reviewed 2026-10-25 · Reference model data · Research notes